AI Paper Atlas

Research map / Image generation / Fast diffusion sampling

Di[M]O: Distilling Masked Diffusion Models into One-step Generator

Yuanzhi Zhu, Xi Wang, Stéphane Lathuilière, Vicky Kalogeiton

ICCV 2025 · Image generation › Fast diffusion sampling · approach: Training-free diffusion acceleration · 19 citations (OpenAlex)

Read the paperCheck your idea against 137,217 papers

AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
distill the MaskGit teacher into a one-step generator with Di[M]OMaskGit · ImageNet-256FID: 16-step teacher 6.60; Di[M]O 1 step 6.91Improved
one-step Di[M]O instead of 4-step teacher samplingMaskGit · ImageNet-256FID: 10.73 -> 6.91Improved
distill Meissonic into one-step generator with Di[M]OMeissonic · HPSv2 averagedHPS score: 4-step teacher 24.66 -> 28.11Improved
Di[M]O one step versus teacher with 48 stepsMeissonic · GenEval overalloverall score: 0.54 -> 0.43Hurt
Di[M]O one step versus other one-step diffusion generatorsDi[M]O · HPSv2 averagedHPS score: 27.48 -> 28.11Improved
set initial mask ratio r_init = 1 (all masked tokens) (ablation)Di[M]O · ImageNet-256FID / training stabilityHurt
set initial mask ratio near 0 (all random tokens) (ablation)Di[M]O · ImageNet-256FID / training stabilityHurt
use hybrid initialization with r_init = 0.6 (ablation)Di[M]O · ImageNet-256FIDImproved
reduce Jeffrey coefficient beta, down to -0.2 (ablation)Di[M]O · ImageNet-256FIDImproved
add Gaussian perturbation on initial token embeddings (ablation)Di[M]O · ImageNet-256FIDImproved

Closest papers by meaning

Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy